Remote Sensing Of Vegetation Phenology In Snow-covered Area Of Northern Hemisphere

As an important phenological characteristic,vegetation green-up date(GUD)is usually estimated by the satellite-based vegetation indices such as the Normalized Difference vegetation index(NDVI).However,due to the influence of spring snowmelt in the snow-covered area,the GUD extracted from the NDVI time series has a large error.Recently,there developed two nominally snow-free vegetation indices for GUD detection,i.e.,the normalized difference phenology index(NDPI)and the normalized difference greenness index(NDGI).Unfortunately,these indices were only tested at several field phenological observation and carbon flux sites,and the accuracy on a wide range of spatial scales lacked comprehensive evaluation,greatly limiting the two vegetation indices on a global scale.In this study,the NDVI,NDPI and NDGI vegetation indices were used to estimate the GUD in the northern hemisphere snow cover area(north latitude above 40° and Qinghai-Tibet Plateau).Firstly,the uncertainty of vegetation returning period of three vegetation indices was quantified in the middle and high latitudes of the northern hemisphere.Then,the distribution of GUD and the interannual rate of change at different spatial resolutions were compared in the Qinghai-Tibet Plateau.The result showed that compared with NDVI,the uncertainties of NDPI and NDGI in the vegetation returning period from 40° to 55° north latitude are small,which can improve the estimation accuracy of GUD,but the three vegetation indices in the region above 55° north latitude show greater uncertainty.In the Tibet Plateau,the GUD distribution and the interannual rate of change reflected by the three vegetation indices are basically the same,and GUD is affected as the resolution increases.Furthermore,selecting which vegetation index to estimate GUD depends on the specific regions.GUD estimations from NDGI are more reliable(i.e.,smaller uncertainty)in the arid and semi-arid grasslands(e.g.,Mongolia and Central Asia grasslands),while NDPI performs slightly better than NDGI in American prairie.In the central and west Europe,Reliable GUD estimations from NDPI and NDGI were only acquired in those years without snowfall.It will be more accurate to select NDPI to calculate GUD in Tibet Plateau.We expect that this study provides the guidelines for phenology detections,particularly using newly developed snow-free vegetation indices,in the seasonal snow-covered areas.